This thesis deals with analysis of the steady state power flow in the Swedish railway electrical system. Finding the steady state of the system means that all voltages and power flows in the system are known.  The trains are supplied from the public grid using frequency converters. This is the major difference compared with ordinary power systems fed from tur­bine-driven generators. The output voltage phase shift function for the con­verters is important in the mathematical formulation of the problem. In the system, the locomotives are almost entirely of the Re-type, having a rectifier and d.c.-motors. The fundamental frequency reactive power need is thereby dependent on the train speed, wheel power and line voltage. In this thesis, the converters and thyristor locomotives are modelled and included in the load flow formulation. The method is shown to have good convergence properties. An approximate and fast method for the calculation of the power injections from converters is presented. A necessary condition for this approximate method to be valid is that the converters are strongly interconnected. The construction of a new 130 kV line brings the converters close to each other electrically. Energy input from each of the converters in­the system can be simulated calculating repeated load flows. Measurements in a real system have verified the applicability of the proposed load flow method. Examples of areas where the method can be used are sys­tem studies and optimal control of the system. The locomotives act as sources of harmonic currents. In an extended load flow formulation this is taken into account. The fundamental idea behind the method is that harmonic powers caused by the thyristor bridges at first are transferred as fundamental frequency active power from the sinusoidal volt­age sources. The method thus gives the fundamental frequency powers injected from the sinusoidal voltage sources to the network. Comparisons with time simulations with detailed models show good agreement.

In this thesis, we investigate the problem of Unmanned Aerial Vehicle (UAV) path planning in search or surveillance mission, when some a priori information about the targets and the environment is available. A search operation that utilizes the available a priori information about the initial location of the targets, terrain data, and information from reasonable assumptions about the targets movement can in average perform better than a uniform search that does not incorporate this information. This thesis provides a simulation-based framework to address this type of problem. Search operations are generally dynamic and should be modified during the mission due to new reports from other sources, new sensor observations, and/or changes in the environment, therefore a Symbiotic Simulation method that employs the latest data is suggested. All available information is continuously fused using Particle Filtering to yield an updated picture of the probability density of the target. This estimation is used periodically to run a set of what-if simulations to determine which UAV path is most promising. From a set of different UAV paths the one that decreases the uncertainty about the location of the target is preferable. Hence, the expectation of information entropy is used as a measure for comparing different courses of action of the UAV. The suggested framework is applied to a test case scenario involving a single UAV searching for a single target moving on a road network. The performance of the Symbiotic Simulation search method is compared with an off-line simulation and an exhaustive search method using a simulation tool developed for this purpose. The off-line simulation differs from the Symbiotic Simulation search method in that in the former case the what-if simulations are conducted before the start of the mission. In the exhaustive search method the UAV searches the entire road network. The Symbiotic Simulation shows a higher performance and detects the target in the considerably shorter time than the other two methods. Furthermore, the detection time of the Symbiotic Simulation is compared with the detection time when the UAV has the exact information about the initial location of the target, its velocity and its path. This value provides a lower bound for the optimal solution and gives another indication about the performance of the Symbiotic Simulation. This comparison also suggests that the Symbiotic Simulation in many cases achieves a “near” optimal performance. / QC 20101110

The aim of the project is to suggest an investment planning programwhere the welfare of the society is to be maximized. In order to beable to decide on a wise investment plan, one needs to know theconsequences of different choices of power system configurations.Therefore the impacts of different future traffic demands are ofinterest for a railway power system owner.Since investments are supposed to last a long time, their futureusage has to be considered. Moreover, the lead times of investmentscan be of considerable duration lengths. Because of the uncertaintyof the future, deterministic case studies might not be suitable andthen a large number of outcomes are to be studied, probable outcomesas well as outcomes with a high level of impact.In order to be able to make a valid long-term investment analysis ofthe railway power supply system, one needs to use proper railwaypower supply models and methods. The aim of this thesis is topresent a stable modeling and methodological basis for the cominginvestment planning phase of this PhD research project. The focus isset on studying the consequences of a railway power supply systemwhich is too weak.The thesis contains an overview of models of some electrical andmechanical relations important for electric traction systems. Someof these models are further developed, and some are modified forimproved computational properties. A flexible electric tractionsystem simulator based on the above mentioned models has beendeveloped and the applied methods and resulting abilities arepresented.The main scientific contribution of this thesis is that a fast andapproximative neural network model, which calculates some importantaggregated results of the interaction between the railway powersystem and the train traffic, has been developed. This approximativemodel was developed in order to reduce computation times. Reductionof computation times is very important when a huge number ofoutcomes are studied. A complete simulation of a train power systemin operation takes a long time, often not less than about a tenth ofthe simulated traffic time. The neural network is trained with someselected aggregated results extracted from a wide set of railwayoperation simulation cases. The choices of network inputs andoutputs are motivated in the thesis. The performance of thesimulator as well as the approximator are visualized in casestudies. / QC 20101101

One of the main challenges posed by 4G wireless communication systems is achieving flexible, programmable multi-standard radio transceivers with maximum hardware share amongst different standards at a minimum power consumption. Evaluating the feasibility and performance of different multi-standard/multi-band radio solutions at an early stage, i.e. system level, is key for succeeding in surmounting this challenge. This entails formulation of the transceiver budget for several RF architectures and frequency plans with different degrees of hardware sharing. This task is complicated by the fact that transceiver blocks can have different implementations that lead to different performances. The tools that are available for use at present have only analysis capabilities or address only one standard and/or receiver architecture at a time. In the belief that a new approach to this problem is necessary, the work that has led to this thesis proposes a novel methodology that automates the design-space exploration of integrated multi-standard wireless radio receivers. This methodology has been implemented in a multi-standard RF Transceiver Architecture Comparison Tool, TACT. TACT helps surmounting many of the challenges faced by RF system designers targeting multi-standard/multi-band radio receivers. The goal of the algorithms TACT is built upon is to find a multi-standard receiver frequency plan and budget that meets or exceeds the specifications of the addressed wireless standards while keeping the requirements of each of the receiver blocks as relaxed as possible. TACT offers RF engineers a deep insight into the receiver behavior at a very early stage of the design flow. It models the impact of critical circuit non-idealities using a high level of abstraction. This reduces the number of design iterations and, thus, the time-to-market of the solution. The reuse of already available intellectual property (IP) blocks is also considered in TACT, what can result in a significant cost reduction of the receiver implementation. A case study of a WCDMA/WLAN multi-standard receiver designed using TACT is presented in order to illustrate the capabilities of the proposed techniques. / QC 20101129

Stator insulation is a critical to the reliability of electrical generators and motors. It is common industrial practice to use electrical measurement of partial discharges (PD) and some form of current-voltage measurement as part of the condition assessment of stator insulation at maintenance times. Extension of these methods by the recently investigated methods of high-voltage dielectric spectroscopy (HV-DS) and variable frequency phase- resolved partial discharge analysis (VF-PRPDA) may offer r usefully increased information about the condition of a stators insulation while requiring less power from the test voltage source than with conventional power-frequency measurements. HV-DS and VF-PRPDA have independent variables of the amplitude and frequency of a sinusoidal voltage applied to the insulation system. The dependent variables are the smooth currents of HV-DS and the discharge pulse charges of VF-PRPDA; these may be analysed in many ways, typically as complex capacitance, the harmonic spectrum of currents when there is PD activity or other non-linearities in the insulation system, and various measures of PD pulse distribution. The methods provide complementary information and have a common need of a variable frequency high voltage driving source. This makes the simultaneous use of these methods a matter of interest, as further information can be gained without extra time and with a total equipment size and cost smaller than that of both separate systems. In this thesis, results are presented from several directions of work relevant to the application of the low frequency diagnostic methods, HV-DS and VF-PRPDA, to machine insulation. The contribution to measured frequency domain dielectric response from the current into the non-linear stress grading of stator end-windings has been studied from physical and numerical models. As well as the effecects on the dielectric response (complex capacitance), the harmonic spectrum of the current into the grading, and the distribution and waveform of the potential along the grading is shown from the numerical models. The frequency and amplitude dependent response of the stress-grading is of importance due to its significacant contribution to the measured dielectric response of a whole stator and to the harmonic currents due to other non-linear phenomena such as partial discharge currents. Short dielectric response measurements have been made on a complete hydro-generator before and after a period of maintenance, giving a better idea of the practical limitations that time constraints and a large test-object put on possible amplitudes and frequencies for driving the test object. Two new epoxy-mica stator coils have been studied with both DS and VF-PRPDA before and after accelerated thermal aging, as a preliminary step for seeing what changes can be detected electrically and for studying how the PD and DS methods differer in their measurement of PD. / QC 20101125

BAE Systems Hägglunds are using contactors to control the power supply in their vehicle,but a replacer has been developed because the contactors take up a lot of space. The replacer is a self-designed semiconductor switch, which is called Ultra High Power DigitalOutPut or UHPDOP. The UHPDOP is a mosfet controlled switch with three separated channels that is constructed to handle 70A each at 28VDC. There was no accomplished test on the circuit and Hägglunds wanted data on the circuits thermal characteristics and switch characteristics. The purpose of this thesis is to investigate if the UHPDOP is a reasonable replacement forthe contactors that are used today. The aim of this thesis is to collect measurement data of the UHPDOP-circuit to discover possible improvements. Be able to argument for or against the UHPDOP-circuit and be able to take a stand against the accuracy that i is needed in mathematic equations and simulations, for it to give some valued informationat electronic construction. The overall question in this thesis is, how accurate is approximate simulations and mathematic equations compared to measurements, in question about transients and heat development in components. To answer the question simulation, mathematical equations and measurements on transient and heatdevelopment in the circuit has been done. Simulation program LTspice, datasheet, IRcamera and oscilloscope has been used during the work. The result shows that • Mathematic equations with only power development and thermal resistance does not give information that is usable at electronic construction. • Simulation gives good understanding of the transients shape. • The UHPDOP-circuit can handle the heat development and transients that it can be exposed to. But the circuit has a fault construction, the current limiters activating current is too low. Aside from that fault construction, some minor parts can be improved. Nothing important for the function, but it would improve the performance in handling of both heat and transients. / BAE Systems Hägglunds i Örnsköldsvik använder sig av kontaktorer för att styra spänningsförsörjningen i sina fordon, men då dessa kontaktorer tar stor plats har enersättare tagits fram. Ersättaren är en egenkonstruerad halvledarströmbrytare som kallas Ultra High Power Digital OutPut eller UHPDOP. UHPDOP-kretsen är en mosfet styrd strömbrytare med tre separata kanaler som var för sig är konstruerade för att hantera 70A vid 28V. Det fanns inga utförda tester på kretsen och Hägglunds var intresserad av data på kretsens termiska egenskaper och switchegenskaper.  Syftet med detta arbete är att undersöka ifall UHPDOP- kretsen är en rimlig ersättare fördagens kontaktorer. Målet är att ta fram mätdata på UHPDOP- kretsens funktion och beteende för att eventuellt upptäcka förbättringsmöjligheter. Kunna argumentera för eller emot UHPDOP- kretsen som en ersättare för kontaktorerna, samt att kunna ta ställning till den noggrannhet som krävs i en simulering och vid en matematisk beräkning för att de ska ge någon användbar information. Frågeställning för arbetet är, hur väl stämmer approximerade simuleringar och matematiska beräkningar mot mätningar vid konstruktion av halvledarströmbrytarkretsar i frågor gällande transienter och effektutveckling. För att besvara frågeställningen har simuleringar, beräkningar och mätningar av transienter och effektutveckling i kretsen utförts, med hjälp av simuleringsprogrammet LTspice, datablad, värmekamera och oscilloskop. Resultatet visar att • Matematiska beräkningar med enbart termiska resistansen och effektutvecklingen för den enskilda kompentent som parametrar inte ger någon användbar information vid elektronikkonstruktion. • Simuleringar ger en god bild över transienters utseende. • UHPDOP-kretsen kan hantera den värmeutveckling och de transienter som den kan komma att utsättas för. Det finns dock en felkonstruktion vilket är att strömbegränsningens aktiveringsnivå är för låg. Utöver felkonstruktionen så finns det förbättringspotential både inom värmeutveckling och hantering av transienter men inget som är kritiskt för dess funktion.

Historical documents are a valuable source of cultural knowledge and can provide information about previous events, societies, beliefs, and cultures. They can serve as an excellent source for research in various fields including history, literature, linguistics, and anthropology. Their preservation and analysis pose significant challenges due to the unique characteristics of handwritten scripts, the variability, and the document degradation. With the rise of the Deep Learning era, enormous amounts of annotated data are required to train large models that can efficiently perform tasks on unseen data. Nowadays, digital libraries provide high-quality digitized images for analysis and processing of historical documents. However, collecting and annotating the provided data is an expensive task and requires a lot of expertise from historians and the humanities. Hence, generating synthetic data to enhance the performance of Deep Learning frameworks is a common approach in Computer Vision and, specifically in this thesis, in Document Image Analysis and Recognition (DIAR). This thesis focuses on leveraging generative models to facilitate DIAR tasks, focusing on historical and handwritten documents, by generating realistic synthetic images that resemble a real distribution and enhance the training of downstream DIAR tasks. The contributions of the thesis include a systematic literature review, a comparison evaluation, and a developed method for handwriting generation. First, a systematic literature review of existing historical document image datasets, provides summarized information of 65 studies, focusing on different aspects, such as statistics, document type, language, visual, and annotation aspects. The study discusses limitations and promising resources for future research, which refer to the limited dataset size and absence of benchmarks, as well as the lack of standardization in terms of data format and evaluation scheme. A subsequent contribution is the integration of generated data in a historical document font classification task. Semi-synthetic data are generated with the use of DocCreator, an open-source software, from which different document degradation augmentations are used. A conditional Generative Adversarial Network (GAN) is used to generate fully synthetic data conditioned on a specific sample. The data generated by the two methods areintegrated as additional samples in the training of several Convolutional Neural Networks classifiers and the effect in the performance is examined. The final contribution of the thesis introduces a new method for generating styled handwritten text images based on Denoising Diffusion Probabilistic Models (DDPM), which is an unexplored method in DIAR. The method manages to capture stylistic and content characteristics of a standard multi-writer handwriting dataset and achieved an improved performance in enhancing writer identification and handwriting text recognition compared to Generative Adversarial Network (GAN)-based methods. The results demonstrate the potential of the generative method for enabling deep document image analysis and pave the way for further research. As a future direction, this work will aim to progress from generating word images to generating sentence and full document images by conditioning on the content, style, and layout of historical documents. Another future action will be to further extend the proposed method to operate in a few-shot scheme for the writer style condition in order to generate unseen styles. Furthermore, the future work will aim to leverage important features from pre-training with synthetic and real data in order to generalize to historical documents that are a scarce source and adjusting the text encoding parts to different languages and scripts. Finally, the ultimate goal of the future work aims to generate a massive synthetic historical document image database to fill the existing benchmark gap.

Examensarbetet handlade om att konstruera en testutrustning som kunde mäta enhiss-dörrs stängningskraft, dess kinetiska energi samt dess strömförbrukning. Dessaparametrar skulle mätas för att kontrollera att hissdörren följde de krav som bestämts i maskindirektivet.Testutrustningen skulle kunna testa befintliga dörröppnare men också nya modellersom först måste kontrolleras att de följde de krav som var ställda. En hårdvara föratt mäta dessa parametrar togs fram, samt ett interface för användaren där olika värden kunde ställas in. De värden som användaren kunde mata in var främst gränsvärden för att avbryta testet om något gränsvärde skulle överskridas. Programmetkunde också hantera kalibrering av de olika sensorerna som användes.I arbetet fanns en sammanfattning om hur idén gick vidare till en prototyp och tillsist blev en färdig produkt med en komplett användarmanual.Resultatet i rapporten bestod av en komplett hårdvara med kompletterande mjukvara. Hårdvaran kunde mäta stängningskraften, dörrens kinetiska energi samt strömförbrukningen. Mjukvaran kunde spara data från de olika testen samt var kompatibelför att kalibreras för olika typer av dörröppnare. Användarmanualen innehöll all information om hur hård- och mjukvaran fungerade samt hur utrustningen användesför att testa dörröppnare.

Detta arbete undersöker förbättringen av fordonsräkning i parkeringshus genom implementering och testning av en kamerabaserad lösning med djupinlärningsalgoritmer. Arbetet utgår från den nuvarande lösningen som är baserad på ultraljudsgivare och som ofta ger ett felaktigt resultat. Målet är att utveckla och implementera en mer tillförlitlig lösning som dessutom kan integreras med uppdragsgivarens nuvarande styrsystem. Arbetet startar i en undersökning av olika mätmetoder som kan appliceras på problemet, däribland ultraljudsgivare som används idag, IR-givare, magnetsensorer och induktiv slinga. Från denna undersökning kunde den kamerabaserade lösningen identifieras som den mest lovande mätmetoden. Detta bekräftas också av tidigare arbeten och forskning inom området, som också visade på att mycket hög noggrannhet kunde uppnås med denna mätmetod. Med detta underlag kunde en kamerabaserad lösningen väljas och senare implementeras i projektarbetet. Arbetet började med ett initialt experiment där en testuppställning med kameralösningens utrustning monterades och anslöts till det lokala nätverket på uppdragsgivarens kontor. Från detta initiala experiment kunde värdefull information om kameralösningens kommunikationsprotokoll och beteende extraheras, för att sedan används för att bygga en färdig applikation som kunde integreras med uppdragsgivarens befintliga styrsystem. Ett andra experiment utfördes i skarpt läge i parkeringshuset Navet i Umeå där den färdigutvecklade applikationen och kameran kunde anslutas till systemet på plats, för att sedan mäta den nya lösningens noggrannhet. Resultatet visade att den nya kamerabaserade lösningen har en noggrannhet som är mycket hög, speciellt i jämförelse med den nuvarande lösningen. Då den nya lösingen inte kalibrerats på 10 dagar uppmätte den samma antal bilar som faktiskt befann sig i garaget, alltså en felmarginal på 0%, medan den nuvarande lösningen som kalibrerades för tre dagar sedan uppmätte en felmarginal på cirka 65%. Den nya kamerabaserade lösningen kunde också på ett effektivt sätt integreras med befintligt styrsystem i fastigheten som möjliggör full integration med P-ledssystemet. Från resultaten kan man tydligt dra slutsatsen att den kamerabaserade har en mycket hög noggrannhet och tillförlitlighet. Trots de positiva resultaten så utfördes projektet med några avgränsningar, där inga extrema väderförhållanden eller miljöförändringar har simulerats. Trots detta kan man anta att resultaten är tillförlitliga, detta då styrkan i en kameralösning är just dess flexibilitet då endast visuell data används. / This thesis investigates the possibility of improving vehicle counting in parking garages by implementing and testing a camera solution with deep learning algorithms. The thesis addresses the client's current issues with an ultrasonic sensor-based solution, which has yielded poor counting results. The goal of this thesis is to develop and implement a more reliable solution that can also be integrated with the client's control systems. The thesis begins by examining various measurement methods that could solve the problem, including the current ultrasonic sensors, as well as IR sensors, magnetic sensors, and inductive loops. From this examination, it became clear that a camera-based solution, which was subsequently implemented, was the best choice. This was further confirmed by related works and research in the field, indicating that a camera-based solution could achieve very high accuracy. The experiments initially included a preliminary test where the camera equipment was mounted at the client's office and connected to the local network. This initial experiment provided valuable information about the camera's communication protocol and behavior. This information was later used to develop a comprehensive application that could integrate the camera solution with the client's current control systems. A second experiment was then conducted in a real parking garage called Navet in Umeå city center, where the developed application and the camera were connected to the system in the garage and tested. The results from this second experiment show that the new camera-based solution with deep learning algorithms has very high accuracy, especially compared to the current solution used in this parking garage. When the new solution was uncalibrated, its counter exactly matched the real number of vehicles in the parking garage, resulting in a margin of error of 0%. The current solution, on the other hand, had been calibrated three days prior and exhibited a margin of error of 65%. The new camera-based solution can also be efficiently integrated with the control systems used by the client and the parking garage. From these results, we can conclude that the new camera-based solution has very high accuracy and is a reliable measurement method for this problem. Despite the positive results of this thesis, it was conducted with limitations, as no extreme weather conditions or environmental changes were simulated. However, one could assume that the results are valid due to the strengths of cameras in their flexibility, as they use visual data which is not sensitive to weather conditions.

Detta examensarbete utforskar utvecklingen av ett övervakningssystem baserat påLow Power Wide Area Network (LPWAN) teknologi för nödutgångsskyltar. Enkund till Knightec som heter Malux äger det befintliga nödbelysningssystem somexamensarbetet är utformat för. Det befintliga systemet kräver att varje nödutgångsskylt kopplas till en central enhet för statusöverföring. Målet med detta arbete är att skapa ett system som gör det möjligt för nödutgångsskyltarna att direktskicka sin status till en molnbaserad lösning, utan behov av centralenhet, vilketökar systemets effektivitet.  I dagens samhälle ökar antalet internetanslutna enheter i hög takt, år 2025 beräknades antalet anslutna enheter för Internet of Things (IoT) vara upp emot 40miljarder, vilket ställer krav att utveckla tekniker som kan hantera lång räckvidd,låg datahastighet, och låg energiförbrukning. De traditionella överföringsteknikerna som Bluetooth och 2G, 3G, 4G mobilkommunikation erbjuder inte en optimalbalans mellan räckvidd och energiförbrukning som LPWAN-teknologin gör. Arbetet har därför fokuserat på att jämföra och analysera de tre ledande LPWANteknologier Sigfox, LoRa och NB-IoT för att bestämma vilken teknologi som ärmest lämplig att appliceras inom övervakning av nödutgångsskyltar. Den praktiska delen av projektet innefattade design och implementering av MKRWAN 1310 som har stöd för LoRaWAN-protokollet. Systemet integrerades medThe Things Network (TTN) för att hantera dataöverföring till molntjänster somAzure IoT Hub och Cosmos DB för lagring och analys. En viktig del av utvecklingsprocessen var att säkerställa att systemet var energieffektivt och kunde hanteraoch överföra data autonomt utan mänsklig inblandning. Resultaten visade att LoRaWAN-teknologin, med dess långa räckvidd och förmåga till låg energiförbrukning, var idealisk för projektets krav. Systemet lyckadeseffektivt övervaka och rapportera status för nödutgångsskyltar, vilket presenterarpotentialen för framtida implementeringar i större skala för att förbättra säkerheten i byggnader och arbetsmiljöer. Detta examensarbete belyser hur modernIoT-teknik kan utvecklas och förbättra säkerhetsinfrastruktur genom innovativalösningar med hjälp av LPWAN-teknologi. / This thesis explores the development of a monitoring system based on Low PowerWide Area Network (LPWAN) technology for emergency exit signs. A client ofKnightec named Malux owns the existing emergency lighting system that the thesisproject is designed for. The existing system requires each emergency exit sign tobe connected to a central unit for status transmission. The goal of this work isto create a system that enables the emergency exit signs to directly transmittheir status to a cloud-based solution, without the need for a central unit, whichincreases the efficency of the system.  In today’s society, the number of internet-connected devices is constantly increasing. By 2025, it is estimated that the number of connected devices for the Internetof Things (IoT) could reach up to 40 billion, which necessitates the developmentof technologies capable of handling long-range, low-data-rate, and low-power consumption requirements. Traditional transmission technologies such as Bluetoothand 2G, 3G, 4G mobile communication do not offer an optimal balance betweenrange and energy consumption like LPWAN technology does. The work has focusedon comparing and analyzing the three leading LPWAN technologies Sigfox, LoRa,and NB-IoT, to determine which technology that is most suitable for monitoringemergency exit signs. The practical part of the project involved designing and implementing MKR WAN1310 that has support for the LoRaWAN protocol. The system was integrated withThe Things Network (TTN) to manage data transmission to cloud services suchas Azure IoT Hub and Cosmos DB for storage and analysis. One important partof the development process was ensuring that the system was energy-efficient andcould handle and transmit data autonomously without human intervention. The results showed that LoRaWAN technology, with its long range and low energy consumption capabilities, was ideal for the project’s requirements. The systemsuccessfully monitored and reported the status of emergency exit signs, presentingthe potential for future implementations on a larger scale to enhance safety in buildings and work environments. This thesis highlights how modern IoT technologycan be developed and improve safety infrastructure through innovative solutionsusing LPWAN technology.